Top all-time · #1301
Sebastian Raschka
Personal blog · magazine.sebastianraschka.comaidata
Sebastian Raschka writes Ahead of AI, a personal blog at magazine.sebastianraschka.com about ai and data. Since 2023, 53 different Hacker News readers have linked it 128 times, most often “Using Local Coding Agents”. It is still writing: the last post was on 2026-10-24, and you can follow it by feed.
Too few sampled comments (22) to say how often HN recommends it.?
Why HN links it
HN readers link Sebastian Raschka's Ahead of AI magazine for explainers on LLM architectures, transformers, LoRA/DoRA, reasoning and fine-tuning. It is often suggested as a learning resource for LLM topics.
- Recommended for · 5 comments
- Called a good explainer and good introductory piece, told readers to read it, and called the newsletter a great resource.
Summarised by Claude from a sample of 22 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Ahead of AI focuses on machine learning and AI research and is read by more than 200,000 researchers and practitioners who want to stay ahead in a rapidly evolving field. Click to read Ahead of AI, by Sebastian Raschka, PhD, a Substack publication with hundreds of thousands of subscribers.
Most-linked posts?
- Using Local Coding Agents6 people · 6 mentions · first linked 2026-06-27
- Controlling Reasoning Effort in LLMs5 people · 5 mentions · first linked 2026-07-18
- The State of LLMs 2025: Progress, Progress, and Predictions4 people · 4 mentions · first linked 2025-12-30
- A Technical Tour of the DeepSeek Models from V3 to v3.24 people · 4 mentions · first linked 2025-12-03
- Developments in LLM Architectures: KV Sharing, MHC, and Compressed Attention4 people · 4 mentions · first linked 2026-05-16
- Language models for text classification: From bag-of-words to Jev4 people · 4 mentions · first linked 2026-09-29
- Understanding Reasoning LLMs3 people · 3 mentions · first linked 2025-02-06
- GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-23 people · 3 mentions · first linked 2025-08-09
- LLM architecture comparison3 people · 3 mentions · first linked 2025-07-19
- Does it beat LLMs? NN+Gzip method reimplemented and explained step-by-step3 people · 3 mentions · first linked 2023-07-30
What Hacker News says?
This is a good explainer: https://magazine.sebastianraschka.com/p/claude-watermarking
Comment linking /p/claude-watermarking
on “Claude Fable 5.1 and Claude Mythos 5.1”Submitted Building LLMs from the Ground Up: A 3-Hour Coding Workshop as a story · 970 points
HN discussionRead this - https://magazine.sebastianraschka.com/p/classifier-history-a...
Comment linking /p/classifier-history-and-jev
on “Clef: Open-weight decision models, and new RL fine-tuning platform”Submitted Sebastian Raschka's from-scratch video course on building reasoning models as a story · 856 points
HN discussionHere is another take on visualizing transformers from Georgia Tech researchers: https://poloclub.github.io/transformer-explainer/ The Illustrated Transformer: https://jalammar.github.io/illustrated-transformer/ Sebastian Raschka, PhD has a post on the architectures: https://magazine.sebastianraschka.com/p/from-gpt-2-to-gpt-os... This HN comment has numerous resources: https://news.ycombinator.com/item?id=35712334
Comment linking /p/from-gpt-2-to-gpt-oss-analyzing-the
on “LLM Visualization”Submitted GPT-6 Astra, looped transformers, and hidden reasoning as a story · 521 points
HN discussionSebastian Raschka has a good introductory piece on what Looped Transformers usually refer to in practice: https://magazine.sebastianraschka.com/p/gpt-6-astra-looped-t... One of the big points is that looped architectures don't necessarily remove CoT style reasoning. They definitely change it, but the model can still be prompted for this kind of output. It just spends more computation on individual tokens. I think th…
Comment linking /p/gpt-6-astra-looped-transformers-and
on “Hybrid Reasoning Agents: LLM agents that think in latent space, act through text”Submitted GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2 as a story · 490 points
HN discussionHere’s some of those sources that were apparently hard to find: https://openai.com/index/openai-o1-system-card/ https://arxiv.org/html/2410.10630v1 https://magazine.sebastianraschka.com/p/understanding-reason...
Comment linking /p/understanding-reasoning-llms
on “GPT-5.6”Submitted Understanding Reasoning LLMs as a story · 473 points
HN discussionIt's more that "random" people are dumb as bricks (but we've in the name of equality and historic measurement errors decided to forgo that), add to it that AI's have a phenomenal (internet sized) memory makes them far more capable than many people. However, even "dumb" people can often make judgements structures in a way that AI's cannot, it's just that many have such a bad knowledge-base that they cannot build the…
Comment linking /p/understanding-reasoning-llms
on “Amazon has mostly sat out the AI talent war”Submitted LLM architecture comparison as a story · 418 points
HN discussionalso https://magazine.sebastianraschka.com/p/understanding-multim...
Comment linking /p/understanding-multimodal-llms
on “Ask HN: Recommendation for a SWE looking to get up to speed with latest on AI”Submitted Language models for text classification: From bag-of-words to Jev as a story · 214 points
HN discussionit's a generalistic classifier w/training needed tho. as far as I'm concerned that is somehwat novel and very practical. you get an ok classifier without working out training data and whatnot. sure, you can do classic ML experiments, but the closest that comes to mind is auto ml. not sure how far we got there, it's been a minute for me on this front. so I don't agree on the premise that it's "just a classifier". it'…
Comment linking /p/classifier-history-and-jev
on “Jeeves. Reasoning improves Jev-like decision models”Submitted Coding Self-Attention, Multi-Head Attention, Cross-Attention, Causal-Attention as a story · 142 points
HN discussionIf you found this interesting, you should read section 5 of this article to learn about how LLMs can be trained so that they're good at confidence scoring: https://magazine.sebastianraschka.com/i/217935364/51-on-cali... OP says that LLMs were never scored against correctness. But that's an assumption about the optimization process. This is sort of right: - pretraining scores on next token, and says nothing about con…
Comment linking /i/217935364/51-on-calibration
on “Using LLM for confidence scoring is a bad idea”Submitted Ten Noteworthy AI Research Papers of 2023 as a story · 128 points
HN discussionThe other recent improvement suggested for LoRA is DoRA: https://magazine.sebastianraschka.com/p/lora-and-dora-from-s... . It really does seem to strongly outperform LoRA - see also https://www.answer.ai/posts/2024-04-26-fsdp-qdora-llama3.htm...
Comment linking /p/lora-and-dora-from-scratch
on “LoRA+: Efficient Low Rank Adaptation of Large Models”Submitted Implementing Weight-Decomposed Low-Rank Adaptation (DoRA) from Scratch as a story · 96 points
HN discussionThanks for the feedback. Clarifying definitely wouldn't hurt. Added a paragraph and new figure at the top of the DoRA section: https://magazine.sebastianraschka.com/i/141797214/introducin... I haven't tried what you were suggesting, but that sounds actually plausible. Interesting idea!
Comment linking /i/141797214/introducing-dora
on “Implementing Weight-Decomposed Low-Rank Adaptation (DoRA) from Scratch”Submitted Understanding Llama 2 and the New Code Llama LLMs as a story · 30 points
HN discussionNot Sebastian (who I assume is the OP), but his blog/substack is also a great resource https://magazine.sebastianraschka.com/
Comment linking /
on “Developing an LLM: Building, Training, Finetuning (A 1h Video Explainer)”Submitted A Technical Tour of the DeepSeek Models from V3 to v3.2 as a story · 23 points
HN discussionCoincidentally, I just started on Build a Large Language Model (From Scratch), a repo/book/course by Sebastian Raschka [0][1][2]. Maybe it is a good problem to have to have to decide which learning resource to use. [0] https://github.com/rasbt/LLMs-from-scratch [1] https://www.manning.com/books/build-a-large-language-model-f... [2] https://magazine.sebastianraschka.com/p/coding-llms-from-the...
Comment linking /p/coding-llms-from-the-ground-up
on “Train Your Own LLM from Scratch”Submitted AI Research Papers in Jan 2024: Model Merging, Mixtures of Experts, Smaller LLMs as a story · 20 points
HN discussionYou would fine tune a pretrained LLM because those books are written in English. And natural languages are in flux, and the corpuses that describe them are not neutral, so you can impose fairness after the fact. Sebastian Raschka has written some relevant popular articles, like: https://magazine.sebastianraschka.com/p/understanding-large-... https://magazine.sebastianraschka.com/p/finetuning-large-lan...
Comment linking /p/finetuning-large-language-models
on “LM Studio – Discover, download, and run local LLMs”Submitted Finetuning Large Language Models as a story · 10 points
HN discussion> Are they literally adding a hidden system prompt that says "effort level: $level" ? Yes. https://magazine.sebastianraschka.com/p/controlling-reasonin...
Comment linking /p/controlling-reasoning-effort-in-llms
on “Maximizing the value of your Claude Code sessions”Submitted Ten Noteworthy AI Research Papers of 2023 as a story · 9 points
HN discussion
